GitHub Availability Report Highlights Progress and Persistent Challenges

GitHub experienced a significant incident on October 28, 2026, when a credential used by GitHub with an upstream model provider became invalid, causing the provider to reject requests as unauthenticated (ID: 1). This affected several GPT-family models in Copilot, leading to errors in Copilot Chat, the Copilot CLI, and code scanning autofix. About one in 10 requests to the affected models failed, with some features like code quality being more heavily impacted. GitHub responded by replacing the invalid credential and deploying it across environments, allowing requests to recover once the first environment was updated. The incident was resolved after confirming that the new credential had been successfully implemented (ID: 1). [1]

On October 30, 2026, GitHub encountered another incident involving a database connection pool used by many github.com services reaching its limit (ID: 1). This caused requests to hang for up to 10 seconds before failing with a server error. Pull request merges and Git operations over SSH and Git LFS were particularly affected, with about 43% of merge attempts failing. The issue arose because some update requests held connections longer than expected, leading the pool to reach its limit. GitHub declared an incident after observing the failures and confirmed that the connection pool had caused the errors. Engineers verified that database connections had returned to normal, and the incident was resolved once monitoring showed stability (ID: 1). [1]

Despite GitHub's efforts to improve reliability, the October 2026 incidents underscored the need for safer operations, stronger isolation, and validated regional recovery (ID: 1). The GitHub Blog reported that while Azure capacity had been expanded and several reliability improvements had been made, the incidents exposed vulnerabilities that required further mitigation. For example, improvements to connection limits and client behaviors were introduced to reduce the chance of dependency interruptions becoming broader reconnect storms. However, the October 28 and October 30 incidents demonstrated that even with these measures, vulnerabilities in authentication management and database connection handling remained (ID: 1). [1]

GitHub experienced an incident on October 28, 2026, when a credential used by GitHub with an upstream model provider became invalid, causing the provider to reject requests as unauthenticated (ID: 1). This led to errors when using several GPT-family models in Copilot, affecting features such as Copilot Chat, the Copilot CLI, and code scanning autofix (ID: 1). The issue was traced to an invalid credential, which was replaced, and requests began to recover once the new credential was deployed across environments (ID: 1). [1]

On October 30, 2026, GitHub encountered another incident involving a database connection pool used by many github.com services reaching its limit, causing requests to hang and then fail (ID: 1). The issue was primarily observed in pull request merges and Git operations over SSH and Git LFS, with about 43% of merge attempts failing at the worst point (ID: 1). The incident was resolved once the connection pool drained and error rates returned to normal, with engineers confirming that the full connection pool had caused the errors (ID: 1). [1]

Despite GitHub's efforts to improve reliability, the October 2026 incidents underscored the need for safer operations, stronger isolation, and validated regional recovery (ID: 1). GitHub has since implemented several targeted reliability improvements, including changes to how the service that assigns jobs to GitHub-hosted runners handles sudden increases in load and improvements in detecting partial network failures (ID: 1). These changes aim to reduce the likelihood and impact of similar incidents in the future (ID: 1). [1]

GitHub experienced an incident on October 28, 2026, when a credential used by GitHub with an upstream model provider became invalid, causing the provider to reject requests as unauthenticated (ID: 1). This issue affected several GPT-family models in Copilot, leading to errors in Copilot Chat, the Copilot CLI, and other Copilot features (ID: 1). The credential issue was resolved by deploying a new credential, which restored functionality after the first environment was updated (ID: 1). [1]

On October 30, 2026, GitHub encountered an incident involving a database connection pool used by many github.com services reaching its limit, causing requests to hang and then fail (ID: 1). The pool was already running near capacity when some update requests started holding their connections for longer than expected, leading to a surge in errors (ID: 1). Engineers resolved the issue by draining the connection pool and returning error rates to normal (ID: 1). [1]

Despite GitHub's efforts to improve reliability, the October 2026 incidents underscored the need for safer operations, stronger isolation, and validated regional recovery (ID: 1). GitHub added substantial independently scalable Azure capacity in October 2026, but the incidents revealed that capacity improvements must be paired with enhanced reliability measures (ID: 1). These events highlight ongoing challenges in managing large-scale systems and ensuring uninterrupted service (ID: 1). [1]

Editorial note: This article is sourced. Some claims are awaiting further verification.

Sources

GitHub Blog

Prepared with AI assistance from the cited sources.